Spatial Algorithmic Bias in Socio-Economic Clustering of Russian Regions

Decision-making based on complex human-machine algorithms can lead to discrimination of citizens based on gender, race and other grounds. However, in world science there is no idea of algorithmically conditioned discrimination of citizens by their place of residence. This also applies to the adoption of algorithmic decisions on the socio-economic development of regions. Therefore, the purpose of our study was to detect algorithmic bias in the results of socio-economic clustering of Russian regions. To achieve this goal, it was necessary to identify sensitive operations in cluster analysis that could lead to spatial injustice, form an array of articles on socio-economic clustering of subjects (regions) of the Russian Federation, analyze all articles for the possibility of algorithmic bias and identify Russian regions with potentially biased attitudes towards them as a result of clustering. The term ‘spatial algorithmic bias’ is proposed. Using the author’s semantic search algorithm in bibliographic databases, six hundred articles with empirical results of cluster analysis of Russian regions by socio-economic indicators were identified. The characteristics of the identified articles are given. The analysis of all the articles showed that algorithmic bias is most evident in the four operations of the clustering algorithm – deploying a conceptual model into an optimal set of indicators, selecting regions, choosing a way to combine regions into clusters and determining the number of clusters. Examples of discriminated Russian regions are presented for each operation. Three directions of further research are indicated. Practical significance may be associated with the adoption of unbiased decisions on regional socio-economic development based on fair clustering of the Russian Federation’s subjects